BR9804457B1 - scale coding / decoding process and apparatus. - Google Patents

Description

SCALE DECODING / DECODING PROCESS AND APPARATUS

BACKGROUND OF THE INVENTION

1. Field of the Invention

The present invention relates to audio encoding and decoding, and more particularly to a scaled audio encoding / decoding process and apparatus using split bit arithmetic encoding. The present invention is adopted as ISO / IEC JTC1 / SC29 / WG 11 N1903 (ISO / IEC Committee Draft 14496-3 SUBPART 4).

2. Description of Related Art State

MPEG audio standards or the AC-2 / AC-3 process provide almost the same audio quality as a compact disc, with a 64-384 Kbps bit rate that corresponds to one-sixth to one- eighth of the conventional digital encoding rate. For this reason, MPEG audio standards play an important role in storing and transmitting audio signals such as digital audio broadcasting (DAB), internet telephone communication, or audio requesting (AOD).

Research into processes where clean audio quality on its original source can be reproduced at a lower bit rate has been done. One process is an MPEG-2 Advanced Audio Encoding (AAC) authorized as a new international standard. MPEG-2 AAC delivering clean audio quality over original 64 kbps sound has been recommended by experts in the field.

In these conventional techniques, a fixed bitrate is provided in an encoder, and the optimized state suitable for the given bitrate is achieved to then perform quantization and encoding, thereby exhibiting considerably better efficiency. However, with the advent of multimedia technology, there is a growing demand for a codec that has versatility at a low bit rate. One type of demand is a scaled audio codec. The scaled audio codec can build from encoded bit groups to a high bit rate into low bit rate bit groups and then restore only some of them. In doing so, the signals can be restored with reasonable efficiency with only some of the bit groups, showing little performance deterioration due to decreased bit rates when an overload is applied to the system or the performance of the data. a decoder is poor, or at the user's request.

According to generic audio coding techniques such as MPEG-2 AAC standards, a fixed bit rate is provided for a coding device, the optimized state for the given bit rate is then searched for execution. quantization and coding, thereby forming bit groups according to the bit rate. A bit group contains information for a bit rate. In other words, bit rate information is contained in the header of a bit group and a fixed bit rate is used. Thus, a process displaying the best efficiency at a specific bit rate can be used. For example, when a bit group is formed by an encoder at a 64 Kbps bit rate, the best sound quality can be restored by a decoder corresponding to an encoder having a 64 Kbps bit rate. .

According to such processes, bit groups are formed without consideration of other bit rates, but bit rates are formed having an appropriate magnitude for a given bit rate rather than the order of bit groups. At present, if the so formed bit groups are transmitted via a communication network, the bit groups are divided into several slots to be transmitted. When an overload is applied to a transmission channel, or only some of the slots sent from a transmission end are received at a receiving end at a narrow bandwidth of the transmission channel, data cannot be be rebuilt properly. Also since bit clusters are not formed according to their significance, if only a few bit clusters are restored, the quality is severely degraded. The reconstructed audio data makes the sound undesirable to the ear. In the case of a scaled audio codec for solving the problems described above, coding is performed for a base layer and then a difference signal between the original signal and the coded signal is coded in the next accentuation layer (K Brandenburg, et al., "First Ideas on Scalable Audio Coding9 7th AES-Conventional preprint 3924, San Francisco, 1994) and (K. Brandenburg, et al.," Three-Stage Bit Rate Scalable Audio Coding System " , 99th AES-Convention, preprint 4132, New York, 1995) Thus, the more layers there are, the poorer the performance at a high bit rate. however, if the state of the communication channels is deteriorated or if the load applied to the decoder of a receiving terminal is increased, a sound with a low bitrate quality is reproduced. and, the coding process mentioned above is not suitable for practically achieving the scale condition.

SUMMARY OF THE INVENTION

To solve the above problems, it is an object of the present invention to provide an apparatus, scaled digital audio data encoding process, and recording means for recording the encoding process using a bit arithmetic encoding technique. (BSAC), rather than a lost coding module with all other conventional coding modules remaining unchanged.

It is another object of the present invention to provide an apparatus, process of scaling digital audio data, and recording means for recording the encoding process, using a split bit arithmetic encoding technique ( BSAC), rather than a lossless encoding module with all other conventional decoder modules remaining unchanged.

To achieve the first object of the present invention, a scaled audio coding process is provided for encoding audio signals in a settled data group having a base layer and accent layers of a predetermined number, comprising the steps of:

input audio signals for signal processing and quantization thereof for each predetermined coding range;

- and packing the quantized data for bit group generation, wherein the bit group generation step comprises:

- coding of quantized data corresponding to the base layer;

- encoding the quantized data corresponding to the next enhancement layer of the encoded base layer and the remaining uncoded quantized data given a layer size limit and belonging to the encoded layer; and

- sequentially performing the layer coding steps for all bit group forming accent layers, wherein the base bed coding step, the accent layer coding step, and the coding step are sequential. are performed so that the lateral information and the quantized data corresponding to a layer to be encoded are represented by digits of the same predetermined one; and then encoded by arithmetic using a predetermined probabilistic model in the order ranging from MSB sequences to LSB sequences, the lateral information containing scale factors and probabilistic models to be used in the arithmetic coding.

step factor coding step with comprises the steps of:

- obtaining the maximum scale factor; and

- obtaining differences between the maximum scaling factor and the respective scaling factors and making the differences difficult by arithmetic.

When the quantized data is composed of signal data and magnitude data, the coding step comprises the steps of:

- coding by a predetermined coding process of the most significant bit sequences composed of the most significant bits of the magnitude data of the quantized data represented by the same number of bits;

- encoding the signal data corresponding to nonzero data between the most significant difficult bit sequences;

- coding the most significant bit sequences among the non-coded magnitude data between the signal data corresponding to non-zero magnitude data between the bit sequences; and

performing the magnitude data coding step and the signal data coding step at the respective bits of the digital data.

The coding steps are performed by the coupling bits composing the respective bit sequences for the magnitude data and the signal data, in bits units of a predetermined number.

A four-dimensional vector coupled in bit units is divided into two subvectors according to their pre-states in the coding of the respective samples.

The base layer bitrate is 16.kbps and the interface bitrate is 8kbps. To achieve the second object of the present invention, a scaled audio coding apparatus is provided comprising:

- a quantization portion for signal processing input audio signals and quantization thereof for each coding range; and

- a bit wrapping portion for generating band-limiting bit groups for a base layer so that it can be scaled, side coding information corresponding to the base layer, sequential coding of quantized information from the most significant bit sequence to the least significant bit sequence, and from the lowest frequency components to higher frequency components, and coding the side information corresponding to the next layer of accentuation of the layer. base and quantized data for coding in all layers.

The quantization portion comprises:

a frequency / time mapping portion for converting audio input signals from a time domain to signals from a frequency domain;

- a psychoacoustic portion for coupling the signals converted by the predetermined sub-range signals by frequency / time mapping and calculation of a substitution operation limit on each sub-range using a substitution operation phenomenon generated by - the interaction of the respective signals;

- and a quantization portion for quantizing the signals for each predetermined coding range while the quantizing noise of each range is compared to the substitution operation limit.

To achieve the third object of the present invention, a scaled audio decoding process for decoding the encoded audio data to have settled bit rates is provided, comprising the steps of:

- decoding lateral information presenting at least scaling factors and arithmetic coding model information distributed for each track, in order of creation of layers in data groups presenting set bitrates, by analyzing the significance of the bits composing the data groups, from the lower significant bits, use of arithmetic coding models corresponding to the quantized data;

- restoration of decoded scale factors and quantized data in signals showing the original magnitudes; and

A inversely converting quantized signals to signals of a time domain.

The decoding of the scale factors is performed by the steps of:

- decoding the maximum scaling factor in the bit group, arithmetic decoding differences between the maximum scaling factor and the respective scaling factors, and subtracting the differences from the maximum scaling factor.

Also provided is an audio scale decoding apparatus comprising:

- a portion analyzing the bit group for decoding the lateral information presenting at least scale factors and arithmetic model information and quantized data, in order of layer creation in nested bit groups;

- an inverse quantization portion for restoration of decoded scale factors and quantized data in signals showing the original magnitudes; and

a time / frequency mapping portion for inverse conversion of quantized signals from a frequency domain to signals from a time domain.

The invention may be incorporated into a general purpose digital computer which is the operation of a program of a computer use medium, including but not limited to the storage medium such as magnetic storage medium (e.g. flexible microdisks). ROM, ROM hard microdisks, etc.), readable media (eg, CD-ROMs, DVDs, etc.) and conductive waves (eg Internet broadcasts). For example, a means of computer use is provided, with tangible incorporation of an instruction program executed by the machine to perform a scaled audio encoding process for encoding audio signals in a set of settled data having a base layer and accent layers of a predetermined number, the process comprising the steps of:

signal processing of the audio input signals and quantization thereof for each predetermined coding range; and

- packaging the quantized data for bit grouping, wherein the bit grouping step comprises:

- coding of quantized data corresponding to the base layer;

encoding the quantized data corresponding to the next accentuation layer of the coded base layer and the remaining uncoded coded data due to a layer size limit and belonging to the coded layer; and

- sequentially performing the layer coding steps for all bit groups forming accent layers, wherein the base layer coding step, the accent layer coding step and the step sequential coding are performed so that the lateral information and the quantized data corresponding to a layer to be coded are represented by digits of the same predetermined number; and

then encoded by arithmetic using the same predetermined probabilistic model in the order ranging from the MSB sequences to the LSB sequences, the lateral information containing scale factors and probabilistic models to be used in the arithmetic coding.

The scale factor coding step comprises the steps of: - obtaining the maximum scale factor; and

- obtaining the differences between the maximum scale factor and the respective scale factors and their arithmetic coding.

The coding of the information for the probabilistic models is performed by the steps of:

- obtaining the minimum value of the information values of the probabilistic model;

- Obtaining the differences in the minimum probabilistic model information and the respective model information values and arithmetic coding using the probabilistic models listed in Tables 5.5 to 5.9.

A means for computer use is provided, with tangible incorporation of a machine-executable instruction program to perform a scaled audio decoding process for decoding the encoded audio data to have settled bit rates, comprising the steps of:

- decoding the lateral information presenting at least scaling factors and arithmetic coding model information distributed for each band, in order of creation of layers in data groups presenting settled bitrates, by analyzing the significance of the composing bits. the data groups, starting from the significant bits higher than the lower significant bits, using arithmetic coding models corresponding to the quantized data;

restoration of decoded scale factors and quantized data in signals showing the original magnitudes; and

conversely converting the quantized data into signals from a time domain, a recording means capable of reading a program for performing the scaled audio coding process using a computer.

The decoding of the scaling factors is performed by decoding the maximum scaling factor, arithmetic coding of the differences between the maximum scaling factor and the respective scaling factors, and subtracting the differences from the scaling factor. maximum.

The decoding of the arithmetic model indices is performed by decoding the minimum arithmetic model index in the bit grouping, decoding the differences between the minimum index and the respective indices in the lateral information of the respective layers and adding the minimum index. and the differences.

BRIEF DESCRIPTION OF DRAWINGS

The above objects and advantages of the present invention will become more apparent from the detailed description of a preferred embodiment thereof with reference to the accompanying drawings in which:

Figure 1 is a block diagram of a single scale decoding / coding apparatus (codec);

Figure 2 is a block diagram of a coding apparatus according to the present invention;

Figure 3 shows the structure of a bit group according to the present invention;

Figure 4 is a block diagram of a decoding apparatus in accordance with the present invention;

Figure 5 illustrates the arrangement of frequency components for an extended block (window size = 2048); and

Figure 6 illustrates the arrangement of frequency components for a short block (window size = 2048).

DESCRIPTION OF PREFERRED EMBODIMENTS

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings.

The bit groups formed in the present invention are not comprised of a simple bitrate, but are made up of several accent layers based on a base layer. The present invention has good encoding efficiency, i.e. the best performance and is displayed at a fixed bit rate as per conventional coding techniques, and relates to a decoding / encoding process and pairing where the bit rate encoded suitable for the advent of multi-media technology is restored.

Figure 2 is a block diagram of a scaled audio coding apparatus according to the present invention, which largely includes a quantization processor 230 and a bit wrapping portion 240.

The quantization processor 230 for signal processing of audio input signals and quantization thereof for predetermined coding range includes a frequency / time mapping portion 200, a psychoacoustic portion 210 and a portion 220 of quantization.

Frequency / time mapping portion 200 converts the input and audio signals from a time domain to signals from a frequency domain. A difference observed between signal characteristics by the human ear is not widely temporal. However, according to human psychoacoustic models, a large difference is produced for each range. Thus, the compression efficiency can be enhanced by the breakdown of different quantization bits depending on the frequency bands.

The psychoacoustic portion 210 couples the converted signals through subband signals determined by the frequency / time mapping portion 200 and calculates a substitution operating limit on each subband using a frequency operation phenomenon. substitution generated by the interaction with the respective signals. The substitution operation phenomenon is a phenomenon where an audio signal (sound) is inaudible due to another signal. For example, when a train passes through a train station, a person cannot hear his neighbor's voice during a low-voice conversation due to the noise caused by the train.

The quantization portion 220 quantizes the signals for each predetermined coding range so that the quantization noise of each band becomes less than the limit for substitution operation. In other words, the frequency signals of each range are applied for scale quantization so that the magnitude of the quantization noise of each range is less than the substitution operating limit so as to be imperceptible. Quantization is performed so that the NMR (Noise Replacement Ratio) value, which is a ratio of the substitution operating limit calculated by the psychoacoustic portion 210 to the noise generated in each range, is less than or equal to 0 dB means that the substitution operation limit is higher than the quantization noise. In other words, quantization noise is not audible.

Bit wrapping portion 240 encodes the side information and the quantized data corresponding to a base layer having the lowest bit rate, successively encodes the side information and the quantized data corresponding to the next accentuation layer of the bed. - from base, and performs this procedure for all layers to generate data clusters. At present, lateral information includes scale factors and probabilistic model information to be used in arithmetic coding.

The coding of the quantized data of the respective layers is performed by the step of dividing each quantized data into bit units by representing the quantified data as binary data comprised of bits of the same predetermined number, and encoding the sequentially divided bit data. from the most significant bit sequence to the least significant bit sequence, using a predetermined probabilistic model. When the digital data is composed of a signal data and magnitude data, the bit wrapping portion 240 collects each of the magnitude data for the bits having the same level of significance between the split bit data, encodes the data into bits. magnitude, and then encode the uncoded signal data between the signal data corresponding to non-zero magnitude data between the encoded magnitude data.

At present, the coding procedure for the data signal and magnitude data is performed in sequence from the MSBs to the lowest significant bits.

The bit clusters formed by the coding apparatus having the configuration mentioned above have a seated structure in which the bit clusters of lower bitrate layers are contained in those higher bitrate layers, as shown in Figure 3. Conventionally, the side information is encoded first and then the remaining information is encoded to form bit groups. However, in the present invention, as shown in Figure 3, the side information for each accent layer is coded separately.

Further, although all quantized data are sequentially encoded in sample units, in the present invention the quantized data is represented by the binary data and is encoded from the MSB sequence of the binary data to form groupings of bits in the binary data. inside of distributed bits.

The operation of the coding apparatus will now be described. Audio input signals are encoded and generated as bit groups. For this purpose, the input signals are converted to frequency domain signals via the Modified Discrete Cosine Transform (MDCT) at frequency / time mapping portion 200. The psychoacoustic portion 210 couples the frequency signals through appropriate subbands to obtain a replacement operating limit. The quantization portion 220 performs scalar quantization such that the magnitude of the quantization noise of each scale factor range is less than the replacement operation limit, which is audible but not noticeable within the bits distributed. If quantization is performed by fulfilling such conditions, the scaling factors for the respective scaling factor ranges and quantized frequency values are generated.

Generally, in view of human psychoacoustics, nearby frequency components can easily be perceived at a lower frequency. However, as the frequency increases, the range of perceived frequencies becomes wider. The bandwidths of the scale factor ranges increase as the frequency ranges become higher. However, to facilitate coding, scale factor ranges of which the bandwidth is not constant are not used for coding, coding bands where the bandwidth is constant are used instead. The coding ranges include 32 quantized frequency coefficient values.

1. Coding of scale factors

For scale factor compression, an arithmetic coding process is used. For this purpose, the maximum scaling factor (max_ scalefactor) is obtained first. Then, the differences between the respective scale factors and the maximum scale factor are obtained and the differences coded by arithmetic. Four models are used in the arithmetic coding of differences between scale factors. The four models are shown in Tables 5.1 through 5.4. Information for the models is stored in a scalefactor model. TABLE 5.1

<table> table see original document page 24 </column> </row> <table>

TABLE 5.2

<table> table see original document page 24 </column> </row> <table>

TABLE 5.3

Differential Scale Factor Arithmetic Model 3

<table> table see original document page 24 </column> </row> <table> TABLE 5.4

Arithmetic Model 4 Differential Scale Factor

<table> table see original document page 25 </column> </row> <table>

2. Arithmetic coding model index coding

Each coding range includes 32 frequency components. The 32 quantized frequency coefficients are coded by arithmetic. Then, a template to be used for arithmetic coding for each coding range is decided, and the information is stored in the Arithmetic Coding Model Index (ArModel). For ArModel compression, an arithmetic coding process is used. For this purpose, the minimum ArModel index (min_ArModel) is first obtained. Then the differences between the respective ArModel indices and the minimum ArModel index are obtained and then the differences are coded by arithmetic. At present, four models are used in the arithmetic coding of differences. The four models are shown in Tables 5.5 through 5.8. Information for the model used in arithmetic coding is stored in an ArModel model.

TABLE 5.5

<table> table see original document page 26 </column> </row> <table> TABLE 5.7

ArModel 3 Arithmetic Differential

<table> table see original document page 27 </column> </row> <table>

TABLE 5.8

Arithmetic Model 4 Differential ArModel <table> table see original document page 27 </column> </row> <table>

1.3 Split-bit coding of quantized frequency components

In general terms, the importance of MSB (most significant bit) is much greater than that of LSB (least significant bit). However, according to the conventional process, coding is performed regardless of importance.

Thus, if only a few bit groups of master bit groups among all bit groups are to be used, the information is less important than the information included in the unused late bit groups and considerably included in the main bit groups. .

For the above reason, in the present invention, the quantized signals of the respective bands are encoded in sequence from the MSBs to the LSBs. In other words, the respective quantized signals are represented by binary notation, and the quantized values of the respective frequency components are sequentially processed in bit units from low frequency components to high frequency components.

First, the MSBs of the respective frequency components are obtained, and then the highest significant bits are encoded in sequence by one bit, up to the LSBs. In this way, the most important information is coded first so that the bit groups are generated from the beginning. The signal values of the quantized values are stored separately and their absolute values are taken to obtain data represented as positive values. The quantized frequency data is divided into bit units and then sequenced from MSB to LSB. The split bit data is reconstructed by four-dimensional vectors. Hence it is assumed that the 8 quantized values each having 4 bits are represented by the binary notation as follows: <table> table see original document page 29 </column> </row> <table>

Conventionally, 1001 of the lowest frequency component is encoded first and then 1000, 0101, 0010 are sequentially encoded (ie, horizontally for each frequency component). However, according to the present invention, 1, MSB sequences composed of the lowest frequency component MSB, and 0, 1, 0, 0, ... MSBs of other frequency components, are obtained and then processed in sequence, being coupled by several bits. For example, in the case of 4-bit encoding, 1010 is encoded and then 0000e encoded. If the coding of the MSBs is complete, the next higher significant bit sequences are obtained and then encoded in the order of 0001,000, ..., up to the LSBs.

The respective four-dimensional vectors coupled in four-bit units are subdivided into two subvectors according to their states. Both sub-users are encoded by an effective lossless encoding such as arithmetic encoding. For this purpose, the model to be used in arithmetic coding for each coding range is decided. This information is stored in ArModel. The respective arithmetic coding models are composed of several low order models. Sub vectors are encoded using one of the low-order models. Low-order models are classified according to the dimensions of the subvector to be encoded, the importance of a vector is decided by the bit position of the vector to be encoded. In other words, according to whether split bit information is for MSB, the following MSB, or LSB, the importance of a vector differs. MSB has the highest importance and LSB has the lowest significance. The coding status values of the respective samples are renewed as the coding of the vehicle progresses from MSB to LSB. First, the coding state value is initialized to zero. Then, when a nonzero bit value is found, the encoding state value becomes 1.

1.4 Signal Bit Encoding

Basically, the coding of a signal bit is performed sequentially from an MSB sequence to an LSB sequence, where the coding of the frequency component data whose signal bit is coded is reserved and the coding of the one whose signal bit is not. is encoded and performed first. After the bit signals of all frequency matching components are encoded in such a manner, the coding of reserved frequency component data is performed in the order of the highest significant bit sequences.

This will be described in more detail. Referring to the above example, the '1010, 0000' sequences are both encoded because their signal bits have not been previously encoded, ie there is no need to reserve encoding.

Then, the following higher significant bit sequences are encoded below '0001, 0000'. At present, for 0001, the first and third 0 are not coded because the signal bits are already coded in the MSBs, and then the second and fourth Oel bits are coded.

At present, since there is no 1 between the upper bits, the signal bit for the frequency component of the fourth bit 1 is encoded. For 0000, since there are no coded signal bits between the upper bits, these four bits are all coded. In this way, the signal bits are encoded to the LSBs, and then the remaining uncoded information is sequenced from the highest significant bits.

1.5 Formation of scaled bit groups

In the following, the structure of the bit groups formed in the present invention will be described. When representing the respective frequency component values in binary digits, the MSBs are encoded first in the base layer, the significant bits that follow are then encoded in the following accent layer, and the LSBs are finally encoded in the base layer. top layer. In other words, in the base layer, only the contour of all frequency components is coded. Then, as the bit rate grows, more detailed information can be expressed. Since detailed information data values are encoded according to increasing bit rates, layer enhancement and audio quality can be improved.

The process for forming scaled bit clusters using such represented data will now be described. First, base layer bit groups are formed. Then the lateral information to be used for the base layer is coded. Side information includes scaling factor information for scaling factor ranges and Arithmetic Coding Model Indexes for each coding range. If the lateral information coding is completed, the information for the quantized values is coded in sequence from the MSBs to the LSBs, and from the low frequency components to the high frequency components. If the distributed bits of a certain range are smaller than those of the range currently being encoded, encoding is not performed. When the distributed bits of the range equal those of the range currently being encoded, encoding is performed. In other words, coding is performed within a predetermined range limit.

The reason for the range limit will be described below. If there is no band limit in the coding signals of the respective accent layers, coding is performed from band independent MSBs. Therefore, unpleasant ear sound can be generated due to being turned on and off when the restoration signals

of the layers have low bit rates. Thus, it is advisable to restrict the bands according to bit rates. Also, if the strips are restricted to the respective accent layers, the complexity decoder for the respective accent layers is reduced. Therefore, both quality and complexity scales can be sustained. After the base layer is encoded, the side information and the quantized value of the audio data for the next accent layer are encoded. In this way data from all layers is encoded. The encoded information is collected all at once to form bit groups. Figure 4 is a block diagram of the decoding apparatus including a bit group analysis portion 400, a reverse quantization portion 410, and a time / frequency mapping portion 420.

The bit group analysis portion 400 decodes the side information presenting at least scale factors and arithmetic coding models, and quantized split-bit data, in the order of generation of bit groups having a seated structure. The decoded data is restored as a time region signal through a processing module in the form of a conventional audio algorithm such as AAC standards. First, the inverse quantization portion 410 restores the decoded scale factor and the quantized data in signals showing the original magnitudes. Time / Frequency Mapping portion 420 converts inversely quantized signals to time domain signals for playback.

Then the operation of the decoding device will be described. The order of the decoding bit groups generated by the coding apparatus is exactly the reverse of the coding order. First, the information for the base layer is decoded. The decoding process will be briefly described. First, the commonly used information for all layers, i.e., the header information stored in the front bit group is first decoded.

The lateral information used in the base layer includes scale factors and arithmetic model indices for the bands distributed in the base layer. Thus, scale factors and arithmetic model indices are decoded. The bits distributed for each coding range can be known by the indices of the decoded arithmetic model. Among the distributed bits, the maximum value is obtained. The quantized values in the bit groups are decoded in sequence from the MSBs to the LSBs7 and from the low frequency components to the high frequency components, as in the coding process. If the distributed bit of a certain range is less than what is currently being decoded, decoding is not performed. When the distributed bit of a certain range becomes equal to the one being decoded at the moment, the decoding is started.

After the decoding of the distributed bit groups to a base layer is completed, the side information and quantized values of the audio data for the next accent layer are decoded. In this way, data from all layers can be decoded.

The data quantized through the decoding process is restored in the form of the original signals through the inverse quantization portion 410 and the time / frequency mapping portion 420 shown in Figure 4 in the reverse order. coding.

Now, a preferred embodiment of the invention will be described. The present invention is adaptable to the AAC standards base structure and implements a scaled digital audio data encoder. In other words, in the present invention, although the basic modules used in the AAC decoding / coding standard are employed, only the lossless coding module is replaced with the split bit coding process. Accordingly, the bit groups formed in the encoder according to the present invention are different from those formed in the AAC technique.

In the present invention, information for only one bit rate is not encoded within a bit group, but information for the bit rates of various accent layers is encoded within a bit group with a nested structure as shown. in Figure 3, in order of order from the most important signal components to the least important signal components.

Using the bit groups thus formed, bit groups having a low bit rate can be formed by simply rearranging the low bit rate bit groups contained in the highest bit group by request. according to the state of the transmission channels. In other words, the bit groups formed by the coding apparatus on a real time basis, or bit groups stored on a medium, may be rearranged to suit the desired bit rate by a user request. , to be then transmitted. Also, if the user's hardware performance is poor or if the user wants to reduce the complexity of a decoder, even with appropriate bit groups, only a few bit groups can be restored, thereby controlling complexity.

For example, in forming a scaled bit grouping, the bit rate of a base layer is 16 Kb-ps, that of the top layer is 64 Kbps, and the respective accent layers have a range. bit rate of 8 Kbps, that is, the bit group has 7 layers of 16, 24, 32, 40, 48, 56 and 64 Kbps. The accent layers are defined as shown in Table 2.1. Since the bit group formed by the coding apparatus has a seated structure as shown in Figure 3, the 64 Kbps top layer bit group contains the bit groups of the respective accent layers ( 16, 24, 32, 40, 48, 56 and 64 Kbps). If the user requests data to the top layer, the bit group of the top layer is transmitted without any processing to it. Also, if another user requests data to the base layer (corresponding to 16 Kbps), only the main bit groups are simply transmitted.

TABLE 2.1

Bit rate for each layer (8 Kbps range)

<table> table see original document page 38 </column> </row> <table>

Table 2.2

Range limit on each layer for short windows (8 kbps ranges

<table> table see original document page 38 </column> </row> <table> TABLE 2.3

Range limit on each layer for large windows (8 Kbps range)

<table> table see original document page 39 </column> </row> <table>

TABLE 2.4

Bits available for each channel in each layer (8 Kbps range)

<table> table see original document page 39 </column> </row> <table> TABLE 2.5

New minimum scaling factor range added for each layer for short windows (8 Kbps range)

<table> table see original document page 40 </column> </row> <table>

TABLE 2.6

New minimum scaling factor range added for each layer for large windows (8 Kbps ranges)

<table> table see original document page 40 </column> </row> <table> Alternatively, accent layers can be constructed at intervals. The bit rate of a base layer is 16 Kbps, that of the top layer is 64 Kbps, and each accent layer has a bit rate range of 1 Kbps. The respective accent layers are constructed as shown in Table 311. Therefore, fine-grained scaling can be implemented, ie scaled bit groups are formed in a bit rate range of 1 Kbps from 16 Kbps to 64 Kbps.

TABLE 3.1

Bitrate for each layer (1 Kbps range)

<table> table see original document page 41 </column> </row> <table> <table> table see original document page 42 </column> </row> <table>

TABLE 3.2

Range limit was each layer for short windows (1 Kbps range)

<table> table see original document page 42 </column> </row> <table> <table> table see original document page 43 </column> </row> <table>

TABLE 3.3

Range limit on each layer for large windows (1 Kbps range)

<table> table see original document page 43 </column> </row> <table> <table> table see original document page 44 </column> </row> <table>

TABLE 3.4

Available bits per channel in each layer (1 Kbps range)

<table> table see original document page 44 </column> </row> <table> TABLE 3.5

New lowest scaling factor range to add on each layer for short windows (1 Kbps range)

<table> table see original document page 45 </column> </row> <table> TABLE 3: 6

New lower scaling factor range to add on each tier for large windows (1 Kbps range)

<table> table see original document page 46 </column> </row> <table>

The respective layers have limited bandwidths according to bit rates. If an 8 Kbps range scale is stressed, bandwidths are limited as shown in Tables 2.2 and 2.3. In the case of 1 Kbps range, the bandwidths are limited as shown in Tables 3.2 and 3.3.

The input data is a 48 KHz sampled PCM data, and the magnitude of a frame is 1024. The number of bits usable for a frame for a 64 Kbps bit rate is 1365.3333 (= 64000 bits / sec * ( Similarly, the available bit sizes for a frame can be calculated according to the respective bit rates. The calculated numbers of bits available for a frame are shown in Table 2.4 for 8. Kbps, and Table 3.4 for 1 Kbps.

2 .1. Coding Procedure

The integral coding procedure is the same as that described in the international MPEG-2 ACC standards, and the split bit coding proposed in the present invention is adopted as lossless coding.

2.1.1. Psychoacoustic Portion

Prior to quantization, using a psychoacoustic model, the block type of a frame currently being processed (extended, start, short, or stop), the SMR values of the respective processing ranges, group information, a short block and temporarily delayed PCM data for frequency / time synchronization with the psychoacoustic model are first generated from the input data, and transmitted to a frequency / time mapping portion. Model 2 ISO / IEC 11172-3 is used for calculating the psychoacoustic model (MPEG Committee IS0 / IEC / JTCl / SC29 / WG11, Information Technology-Coding of Moving Pictures and Associated Audio for Data Storage Media to About 1.5 Mbitts-Part 3: Audio, ISO./OEC 11172-3, 1993).

2.1.2. Frequency / Time Mapping Portion

Frequency / time mapping defined in international standards MPEG-2 AAC is used. The frequency / time mapping portion converts data from a time domain to data from a frequency domain using MDCT according to block type output using the psychoacoustic model. At this time the block sizes are 2048 and 256 for stop / start / extend blocks and for a short block respectively and MDCT is performed 8 times (MPEG Committee ISO / IEC / JTC1 / SC29 / WG11, MPEG-3 AAC IS 13818-7, 1997). The same procedure as that used in conventional MPEG-2 AAC (MPEG Committee ISO / IEC / JTC1 / SC29 / WG11, ISO / IEC MPEG-2 AAC IS 13818-7) was used hereinafter.

2.1.3. Quantization Portion

Data converted to that of a frequency domain is quantized with increasing scaling factors such that the SNR value of the scaling factor range shown in Tables 1-1 and 1-2 is less than SMR as the output value of the scaling factor. psychoacoustic model. At present, scalar quantization is performed, and the basic scaling factor range is 2 "'" // 4. Quantization is performed so that noticeable noise is minimized. The exact quantization procedure is described in MPEG-2 AAC. At present, the output obtained comprises the quantized data and the scaling factors for the respective scaling factor ranges.

TABLE 1.1

Scale factor range for long blocks

<table> table see original document page 49 </column> </row> <table> <table> table see original document page 50 </column> </row> <table>

TABLE 1.2

Scale factor range for short blocks

<table> table see original document page 50 </column> </row> <table>

2.1.4. Frequency component arrangement

For convenient coding, the frequency components are rearranged. The rearrangement order differs depending on the block types. If an extended block type window is used, the frequency components are arranged in the order of the scale factor ranges, as shown in Figure 5. In Figure 5, sfb indicates scale factor range. If a short block type window is used, each four frequency components from eight blocks are arranged repeatedly in ascending order, as shown in Figure 6. In Figure 6, B indicates 8 blocks, and the digits listed below indicate frequency components in each block.

2.1.5. Bit wrapping portion using split bit arithmetic condition

(BSAC)

Quantized data and scaled factors are formed as nested bit groups.

Bit groups are formed by syntaxes shown in Tables 7.1 to 7.10. TABLE 7.1

Raw data stream () syntax

<table> table see original document page 52 </column> </row> <table>

TABLE 7.2

Raw data block () syntax

<table> table see original document page 52 </column> </row> <table> TABLE 7.3

Single channel element () syntax

<table> table see original document page 53 </column> </row> <table>

TABLE 7.4

Ics info syntax ()

<table> table see original document page 53 </column> </row> <table> TABLE 7.5

Bsac channel stream syntax ()

<table> table see original document page 54 </column> </row> <table>

TABLE 7.6

Bsac data () syntax

<table> table see original document page 54 </column> </row> <table> TABLE 7.7

Bsac stream () syntax

<table> table see original document page 55 </column> </row> <table> <table> table see original document page 56 </column> </row> <table> TABLE 7.10

Bsac_spectral data () syntax

<table> table see original document page 57 </column> </row> <table> <table> table see original document page 58 </column> </row> <table> The main elements of a bit group are elements which can usually be used in conventional ACC, and the novel elements proposed in the present invention are specifically explained. However, the main structure is similar to that of the AAC standards. Next, the elements of a new bit group proposed in the present invention will be described.

Table 7.5 shows the syntax for coding bsac_ channel_stream, where the max_scale factor represents the maximum scaling factors, which is an integer, ie 8 bits.

Table 7.6 shows the syntax for encoding bsac_data; where * frame_lenght 'represents the size of all bit groups for a frame, which is expressed in units of bytes. Also, 'encoded_layer' represents the encoding for the bit group-encoded top layer, which is 3 bits for the 8 Kbps range and 6 bits for the 1 Kbps range, respectively. The information for the accent layer is shown in Tables 2.1 and 3.1. Also, 'scaler factor_model' represents information concerning models to be used in arithmetic coding differences in scale factors.

These models are shown in Table 4.1. The 'min ArModel' represents the minimum value of the arithmetic coding model indices. 'ArModel_model1' represents information regarding models used in arithmetic coding of a difference signal between ArModel and min_ArModel. This information is shown in Table 4.2. TABLE 4.1

Differential Scale Factor Arithmetic Model

<table> table see original document page 60 </column> </row> <table>

TABLE 4.2

Differential ArModel Arithmetic Model

<table> table see original document page 60 </column> </row> <table>

Table 7.9 shows the syntax for encoding bsac_side_info. The information that can be used for all layers is first encoded and then the side information commonly used for the respective accent layers is encoded. 'Acode_scf' represents a keyword obtained by arithmetic coding of scale factors. 'Acode_ArModel' represents a keyword obtained by ArModel arithmetic coding. ArModel is the information that is selected from the models listed in Table 4.3.

TABLE 4.3

BSAC Arithmetic Model Parameters

<table> table see original document page 61 </column> </row> <table> <table> table see original document page 62 </column> </row> <table>

TABLE 6.1

Arithmetic Model 0 BSAC

Distributed Bit = 0 <table> table see original document page 63 </column> </row> <table>

TABLE 6.2

<table> table see original document page 63 </column> </row> <table> TABLE 6.4

Arithmetic Model 4 BSAC

Distributed Bits = 2

<table> table see original document page 64 </column> </row> <table> <table> table see original document page 65 </column> </row> <table>

TABLE 6.5

<table> table see original document page 65 </column> </row> <table> <table> table see original document page 66 </column> </row> <table>

TABLE 6.6

<table> table see original document page 66 </column> </row> <table> <table> table see original document page 67 </column> </row> <table> <table> table see original document page 68 < / column> </row> <table>

TABLE 6.7

<table> table see original document page 68 </column> </row> <table> <table> table see original document page 69 </column> </row> <table> TABLE 6.8

<table> table see original document page 70 </column> </row> <table> <table> table see original document page 71 </column> </row> <table> <table> table see original document page 72 < / column> </row> <table>

TABLE 6.9 "

<table> table see original document page 72 </column> </row> <table> <table> table see original document page 73 </column> </row> <table> <table> table see original document page 74 < / column> </row> <table> <table> table see original document page 75 </column> </row> <table>

TABLE 6.10

<table> table see original document page 75 </column> </row> <table> <table> table see original document page 76 </column> </row> <table> <table> table see original document page 77 < / column> </row> <table> <table> table see original document page 78 </column> </row> <table>

TABLE 6.11

Arithmetic Model 11 BSAC

Distributed Bits (Abit) = 5

<table> table see original document page 78 </column> </row> <table> <table> table see original document page 79 </column> </row> <table> <table> table see original document page 80 < / column> </row> <table> <table> table see original document page 81 </column> </row> <table>

TABLE 6.12

Arithmetic Model 12 ASAC

Identical to arithmetic model 10 BSAC, but with bit distributed = 6 TABLE 6.13

Arithmetic Model 13 ASAC

Identical to arithmetic model 11 BSAC, but with distributed bit = 6.

TABLE 6.14

ASAC Arithmetic Model 14

Identical to arithmetic model 10 BSAC, but with distributed bit = 7.

TABLE 6.15

ASAC Arithmetic Model 15

Identical to arithmetic model 11 BSAC, but with distributed bit = 7.

TABLE 6.16

ASAC Arithmetic Model 16

Same as arithmetic model 10 BSAC, but with distributed bit = 8.

TABLE 6.17 ASAC Arithmetic Model 17

Same as arithmetic model 11 BSAC, but with distributed bit = 8.

TABLE 6.18

ASAC Arithmetic Model 18

Identical to arithmetic model 10 BSAC, but with distributed bit = 9.

TABLE 6.19

Arithmetic Model 19 ASAC

Identical to arithmetic model 11 BSAC, but with distributed bit = 9.

TABLE 6.20

ASAC Arithmetic Model 20

Identical to arithmetic model 10 BSAC, but with distributed bit = 10.

TABLE 6.21

Arithmetic Model 21 ASAC <table> table see original document page 84 </column> </row> <table>

TABLE 6.22

<table> table see original document page 84 </column> </row> <table>

TABLE 6.23

<table> table see original document page 84 </column> </row> <table>

TABLE 6.24

<table> table see original document page 84 </column> </row> <table>

TABLE 6.2 5

<table> table see original document page 84 </column> </row> <table> buido = 12.

TABLE 6.2 6

<table> table see original document page 85 </column> </row> <table>

TABLE 6.2 7

<table> table see original document page 85 </column> </row> <table>

TABLE 6.28

<table> table see original document page 85 </column> </row> <table>

TABLE 6.29

<table> table see original document page 85 </column> </row> <table> buido = 14.

TABLE 6.30

Arithmetic Model 30 ASAC

Identical to arithmetic model 10 BSAC, but with distributed bit = 15.

TABLE 6.31

ASAC Arithmetic Model 31

Identical to arithmetic model 11 BSAC, but with distributed bit = 15.

Table 7.10 presents the syntax for encoding bsac_spectral_data. The lateral information usually used for the respective accent layers, the quantized frequency components are divided into bits using the BSAC technique and then coded by arithmetic. 'Acode_vec0' represents a keyword obtained by arithmetic coding of the first subvector (subvector 0) using the arithmetic model defined as the ArModel value.

'Acode_vecl' represents a keyword obtained by arithmetic coding of the second subvector (subvector 1) using the arithmetic model defined as the ArModel value.

'Acode_sign' represents a keyword obtained by arithmetic coding of the signal bit using the arithmetic model defined in Table 5.9.

TABLE 5.9

<table> table see original document page 87 </column> </row> <table>

While the number of bits used in encoding the respective subvectors is calculated and compared with the number of bits available for the respective emphasis layers, when the bits used are equal to or greater than the available bits, it is again the next accent layer starts.

In the case of an extended block, since the base layer bandwidth is limited to the twenty-first scale factor range, then the scaling factors to the twenty-first factor range are coded. and the arithmetic coding models of the corresponding coding ranges. Bit distribution information is obtained from arithmetic coding models. The maximum value of the distributed bits is obtained from the distributed bit information for each coding range, and coding is performed from the maximum quantization bit value through the coding process mentioned above. Then the following quantized bits are encoded in sequence. If the bits distributed in a certain range are smaller than those in the range being encoded, encoding is not performed.

When the distributed bits of a certain range are equal to those of the range currently being encoded, the range is encoded for the first time. Since the base layer bitrate is 16 Kbps, full bit availability is 336 bits. Thus, the total amount of bits used is continuously calculated and the encoding is terminated at the moment the bit quantity exceeds 336.

After the bit groups for the base layer (16 Kbps) are formed, the bit groups are formed for the next accent layer. Since the limited bandwidths are increased for the higher layers, the scaling factor coding and arithmetic coding models are performed only for the new bands added to the limited bands for the base layer. At the base layer, the uncoded bit data divided for each range and the split bit data of a new added range are encoded from the MSBs in the same manner as at the base layer. When the total bit amount used is greater than the available bit amount, the coding is terminated and the bit grouping of the next accentuation layer is prepared. In this way, bit groups can be generated for the remaining 32, 40, 48, 56 and 64 Kbps layers.

The decoding procedure will now be described.

3.1. Bit group analysis and decoding.

3.1.1. Bsac_channel_stream decoding.

The decoding of bsac_channel_stream is performed in the order of Get max_scale factor for Get ics_info () and Get BSAC data, as shown in Table 7.5.

3.1.2. Bsac_data decoding.

The required side information in frame length decoding, encoded layer, scale factor models, and arithmetic models is decoded in the bit group, as shown in Table 7.6. 3.1.3. Bsac_side_info decoding.

The scaled bit groups formed above have a seated structure. First, the side information for the base layer is separated from the bit group and then decoded. Then the split bit information for the quantized frequency components contained in the base layer bit group is separated from the bit group and then decoded. The same decoding procedure as that for the base layer is applied to other enhancement layers.

1) Decoding of scale factors.

Frequency components are divided into scaling factor ranges with frequency coefficients that are multiples of 4. Each scaling factor range has a scaling factor. The max_scale factor is decoded into an unsigned 8-bit integer. For all scale factors, the differences between the scale factors and the max_scale factor are obtained and then coded by arithmetic. The arithmetic models used in decoding the differences are one of the elements forming the bit groups, and are separated from the already encoded bit groups. The original scaling factors can be restored in the reverse order of the codification procedure.

The following pseudo code describes the decoding process for the scaling factors at the base layer and the other stress layers.

for (g = 0; <g num_window_group; g + + {

para (sfb = layer_sfb [layer]; sfb <layer_sfb [layer + 1]; sfb ++) {

sf [g "|" [sfb ~] = max_scale factor - arithmetic_decoding (); }}.

At present, layer_sfb (layer) is a scaling factor start range for decoding the scaling factors in the respective accent layers, and layer_sfb (layer +1) is a scaling factor ending range.

2) Arithmetic model index decoding.

Frequency components are divided into coding ranges presenting frequency coefficients to be coded without loss. The coding range is a basic unit used for lossless coding. The arithmetic coding model index is information on the models used in the arithmetic decoding / coding of the split bit data of each coding range, indicating which model is used in the arithmetic decoding / coding procedures among the models listed in Table 4.3.

Differences between a shifted value and all arithmetic coding model indices are calculated and then the difference signals are coded by rhythm using the models listed in Table 4.2. At present, among the four models listed in Table 4.2, one model can be used as indicated by the value of Ar Model_model and is stored in the 2-bit bit group. The offset value is a 5-bit min_ArModel value stored in the bit group. The difference signals are coded in the reverse order of the coding procedure and then the difference signals are added to the offset value to restore the coding model indices to rhythm.

The following pseudo code describes the decoding process for the arithmetic and ArModel (cband) coding model indices in the respective stress layers.

to (sfb = layer_sfb (layer); sfb <layer_sfb

(layer + 1); sfb ++)

for (g = 0: g <num_window_group; g ++) {

range = (sfb * num_window_group) + g

for (i = 0; swb_offset (range); i <swb_offset (layer +1); i + = 4) {cband = index = cb (g, i);

if (! decode_cband (ch) (g) (cband)) {

ArModel (g) (cband) = min_ArModel + arithmetic_ decoding ()

decode_cband (ch) (g) (cband) = 1; At present, layer_sfb (layer) is a scale factor start range for decoding the arithmetic coding model indices in the respective accent layers, and layer_sfb (layer +1) is a scale factor end range. The decode_cband (ch) (g) (cband) is an indicative display of whether an arithmetic coding model has been decoded (1) or not decoded (0).

3.1.4. Decoding of the split bit data.

Quantized sequences are formed as split bit sequences. The respective four-dimensional vectors are divided into two subvectors according to their states. For effective compression, the two subvectors are coded by arithmetic in the form of lossless coding. The model to be used for arithmetic coding for each coding range is decided. This information is stored in ArModel.

As shown in Tables 6.1 through 6.31, the respective arithmetic coding models are composed of several low-order models. Subvectors are coded using one of the low-order models. Low-order models are classified according to the size of the subvector to be coded, the importance of a vector, or the coding states of the respective samples. The importance of a vector is decided by the bit position of the vector to be encoded. In other words, according to whether split bit information is for MSB, MSB, next MSB, or LSB, the importance of a vector differs. MSB is of utmost importance and LSB of least importance. The coding state values of the respective samples are renewed as the vector coding progressed from MSB to LSB. In the beginning, the coding state value is initialized to zero. Then, when a nonzero bit value is found, the value of the coding state becomes 1.

The two subvectors are uni to four-dimensional vectors. Subvectors are arithmetically encoded from MSB to LSB, from lower frequency components to higher frequency components. Arithmetic coding model indices used in arithmetic coding are previously stored in the bit group in order of lowest frequency to highest frequency, before transmission of the bit data split for each coding range in units of. coding ranges.

The respective split bit data is encoded by arithmetic to obtain the keyword indices. These indices are restored to the original quantized data and coupled in bits using the pseudo code to follow. The pre_state () is a state indicator of whether the current decoded value is 0 or not. 'Snf' understands the importance of a decoded vector. 0 '1dx0' is a keyword index whose previous state is 0. 0 'idx1' is a keyword index whose previous state is 1. 0 'dec_sample O1 is the decoded data. 0 'start_i' is the line for (i = start_i; i <(start_i + 4); i ++) {

if (pre_state (i)) {

if (idxl & 0 χ 01)

dec_sample (i) | = (1 <<(snf-1))

idxl>> = 1;

}

else {

if (idx0 & 0 χ 01)

dec_sample (i) | = (1 <<(snf-1))

idx0>> = 1;

}

}

While the divided bit data of the quantized frequency components is encoded from MSB to LSB, when the signal bits of non-zero frequency coefficients are encoded by arithmetic. A negative sign bit (-) is represented by 1 and a positive sign bit (+) is represented by 0.

Therefore, if the split bit data is decoded by arithmetic in a decoder and a non-zero bit value decoded by arithmetic is found first, then the signal information in the bit grouping follows, i. Yeah, acode_sign. Bit_sign is decoded by artmetic using this information with the models listed in Table 5.9. If the sign_bit is 1, the signal information is given to the quantized data (y) formed by coupling the separate data as follows.

<formula> formula see original document page 96 </formula>

3.2. Inverse Quantization.

The inverse quantization portion restores the decoded scale factors and the quantized data in signals showing the original magnitudes. The inverse quantization procedure is described in AAC standards.

3.3. Frequency / Time Mapping.

The frequency / time mapping portion inversely converts audio signals from a frequency domain into signals from a time domain to be reproduced by the user. The formula for frequency domain signal mapping in time domain signal is defined in AAC standards. Also, various items such as a mapping related window are also described in the AAC standards. The above-mentioned embodiment of the present invention may be formed as a program executable on a computer. The program can be stored on a recording medium such as a CD-ROM, hard disk, floppy disk or memory. The means of registration is commercially available. The recording medium is obviously within the scope of the present invention.

The present invention may be embodied in a general purpose digital computer running a program from a computer use medium, including but not limited to the storage medium such as magnetic storage medium (e.g. floppy disks, ROM hard disks, etc.), optical readout (eg CD-ROMs, DVDs, etc.) and conductive waves (eg Internet transmissions).

Therefore, the present invention may be embodied as a means of computer use having computer readable program code means incorporated therein for encoding a sequence of digital data of a predetermined number, the means of computer readable program code in the use of computer comprising computer readable program code means for causing a computer to perform signal processing of the audio input signals and quantization of the same for each predetermined coding range, and a means of computer readable program code to cause a computer to perform packetization of quantized data to generate bit clusters, wherein the step of generating a bit group. bits comprises coding of quantized data corresponding to the base layer, coding of quantized data corresponding to the following concentration layer of and coded base and the remaining unencrypted quantized data due to a layer size limit and being part of the coded layer, and sequentially performing the layer coding steps for all bit group formation enhancement layers, in that the base layer coding step, the accent layer coding step and the sequence coding step are performed so that the side information and the quantized data corresponding to a layer to be coded are represented. by digits of the same predetermined number; and then encoded by rhythmics using a predetermined probabilistic model in the order ranging from MSB sequences to LSB sequences, the lateral information containing scale factors and probabilistic models to be used in arithmetic coding. A functional program, code segments, and code used to implement the present invention may be derived by a specialized programmer from the description of the invention contained herein.

According to the present invention, while using conventional audio algorithm such as MPEG-2 AAC standards, only lossless coding portion is modified to enable scaling.

Also, once the audio algorithm is used, the operation required to implement the present invention is simplified.

Since bit groups can be embedded in scales, a bit group can contain multiple bit groups having multiple bit rates. If the present invention is combined with AAC standards, almost the same audio quality can be obtained at the top layer bit rate.

Also, since coding is performed according to the importance of bit quantization, rather than performing coding after processing the difference between the quantized signals of the previous layer and the original signal for each layer, The complexity of the coding apparatus is reduced.

Since a bit group contains multiple bit groups, bit groups for multiple layers can be simply generated and the complexity of a transcoder is reduced.

If the bit rate is reduced due to limited ranges, the complexity of a filter, which is a major source of complex encoding and decoding, is considerably reduced. Consequently, the complexity of a coding and decoding apparatus is reduced.

Also, depending on the performance of the user decoders and the congestion / bandwidth of the transmission channels or at the user's request, bit rates or complexity can be controlled.

To satisfy multiple user requests, flexible bit groups are formed. In other words, upon request from the user, the information for the multi-layer bitrates is combined with a non-overlapping bit grouping, thereby providing bit groups presenting good audio quality. Also, no converter is required between a transmit terminal and a receive terminal. In addition, any state of the transmission channels and various user requests can be accommodated.

Claims (26)

1. SCALE AUDIO ENCODING PROCESS, for encoding the audio signals in a nested data grouping having a base layer and accentuation layers of a predetermined number, comprising the steps of: - audio input signals of signal processing (200), and quantization (220) thereof for each predetermined coding range; and packing (240) the quantized data for bit group generation, wherein the bit group generation step is characterized by comprising: - encoding the quantized data (330-0) corresponding to the base layer; Encoding the quantized data (330-1) corresponding to the next accentuation layer of the encoded base layer and the remaining unencoded quantized data due to a layer size limit and forming part of the encoded layer; and sequentially performing the layer coding steps for all bit group formation enhancement layers, wherein the base layer coding step, the accent layer coding step, and the sequence coding step are performed such that the side information (310-0, 310-1, 310-N) and quantized data (330-0, 330-1, 330-N) corresponding to a layer to be encoded are represented by digits of same predetermined number; and - and then encoded by arithmetic using a predetermined probabilistic model in the order of the MSB sequences to the LSB sequences, in consideration of the bit significance, the lateral information (310-0, 310-1, 310-N) containing factors of scale and probabilistic models to be used in arithmetic coding. Scaling audio coding process according to Claim 1, characterized in that the scaling factor coding step comprises the steps of: - obtaining the maximum scaling factor; and - obtaining the differences between the maximum scaling factor and the respective scaling factors and arithmetic coding of the differences. Process according to Claim 2, characterized in that the probabilistic models listed in Tables 5.1 to 5.4 are used in the arithmetic coding step of the differences. Process according to Claim 1, characterized in that the probabilistic models listed in Tables 6.1 to 6.31 are used in the arithmetic coding step. Process according to Claim 4, characterized in that the information coding for the probabilistic models is performed by the steps of: - obtaining the minimum value of the probabilistic model information values; - obtaining the differences between the minimum probabilistic model information and the respective model information values and the arithmetic coding of the differences. Process according to Claim 5, characterized in that the probabilistic models listed in Tables 5.5 to 5.8 are used in the arithmetic coding step. Process according to Claim 1, characterized in that when the quantized data is composed of the signal data and the magnitude data, the coding step comprises the steps of: - coding by a predetermined coding process of the sequences of most significant bits composed of the most significant bits of the magnitude data of the quantized data represented by the same number of bits; - encoding the signal data corresponding to nonzero data between the most significant bit sequences; - coding the most significant bit sequences among the uncoded magnitude data of the digital data by a predetermined coding process; encoding the non-encoded signal data between the signal data corresponding to non-zero magnitude data between the bit sequences; and performing the magnitude data coding step and the signal data coding step at the respective digital data bits. Process according to Claim 7, characterized in that a probabilistic model has a size of 2 and cumulative frequency values of 8192 and 0 is used in the arithmetic coding step of the signal data. Process according to Claim 7, characterized in that the coding steps are performed by the coupling bits composing the respective bit sequences for the magnitude data and the signal data, in bit units of a predetermined number. . Process according to Claim 9, characterized in that the number of bits is 4. Process according to either of Claims 9 and 10, characterized in that a four-dimensional vector coupled in bit units is divided into two subvectors according to their pre-states in the coding of the respective samples. Process according to Claim 1, characterized in that the base layer bitrate is 16 Kbps and the interface bitrate is 8 Kbps. Process according to Claim 1, characterized in that the bit rate is 16 Kbps and the interface bit rate is 1 Kbps. Process according to Claim 1, characterized in that the header information commonly used for all bands and the lateral information and quantized frequencies required for the respective layer are formed by the split bit information to then be encoded to present a seated structure. Process according to Claim 1, characterized in that the quantization is performed by the steps of: converting (200) the audio input signals of a time domain into signals of a frequency domain; - coupling (210) the signals converted to predefined subband signals — by frequency / time mapping and calculating a substitution operating limit for each subband; and - quantizing (220) the signals for each predetermined encoding range so that the quantizing noise of each range is less than the substitution operating limit. A scaled audio coding apparatus comprising: - a quantization portion (230) for signal processing of the audio input signals and quantization thereof for each coding range, the apparatus being characterized by: - a bit wrapping portion (240) for generation of bit groups by band-limiting to a base layer to provide scaling, side coding information (310-0, 310-1, - 310-N) corresponding to base layer, coding of quantized information (330-0, 330-1, 330-N) in sequence from the most significant bit sequence to the least significant bit sequence, and from lowest frequency components to higher frequency, in consideration of bit significance, and coding of the side information corresponding to the next accentuation layer of the base layer and the quantized data, to perform coding on all layers. Apparatus according to claim 16, characterized in that the quantization portion (230) comprises: - a frequency / time mapping portion (200) for converting audio inputs from a time domain to signals from a frequency domain; - a psychoacoustic portion (210) for coupling the signals converted by the predetermined subband signals by frequency / time mapping and calculating a substitution operating limit for each subband using a substitution operation phenomenon generated by the interaction of the respective signals; and - a quantization portion (220) for quantizing the signals for each predetermined coding range, while the quantizing noise of each range is compared to the substitution operating limit. 18. SCALE AUDIO DECODING PROCESS, for decoding audio data encoded to have settled bit rates, comprising quantized data in signals having the original magnitudes, and converting (420) the quantized signals inversely in time domain signals, the process being characterized by the steps of: - decoding (400) the lateral information (310-0, - 310-1, 310-N) presenting at least scale factors and arithmetic coding of the distributed model information for each band; in order of creation of the layers in the data groups presenting set bit rates, by analyzing the importance of the bits composing the data groups, from the significant bits higher than the lower significant bits, using the arithmetic coding models corresponding to the quantized data. (330-0, 330-1, 330-N); - restoration (410) of the decoded scale factors. Process according to Claim 18, characterized in that the bit groups are decoded into four-dimensional vector units. Process according to Claim 19, characterized in that the four-dimensional vectors are restored from two arithmetically decoded subvectors according to the coding states of the respective samples. Process according to Claim 20, characterized in that the probabilistic models listed in Tables 6.1 to 6.31 are used in the arithmetic decoding step. Process according to Claim 18, characterized in that the decoding (200) of the scaling factors is performed by the steps of: - decoding the maximum scaling factor in the bit group, arithmetic decoding of the differences between the scaling factor and the respective scaling factors, and subtraction of the differences from the maximum scaling factor. Process according to Claim 22, characterized in that the probabilistic models listed in Tables 5.1 to 5.4 are used in the arithmetic decoding step. Process according to claim 18, characterized in that the decoding (400) of the arithmetic model indices is performed by decoding the minimum arithmetic model index in the bit group, decoding the differences between the minimum index and the respective indices in the bit group. lateral information of the respective layers, and adding the minimum index and the differences. Process according to Claim 22, characterized in that the probabilistic models listed in Tables 5.5 to 5.8 are used in the arithmetic decoding step. 26. SCALE AUDIO DECODING EQUIPMENT, comprising quantized signal data presenting the original magnitudes and a frequency / time mapping portion (420) for inversely converting quantized signals from a frequency domain to signals from a time domain, apparatus being characterized by comprising: - a bit group analysis portion (400) for decoding lateral information (310-0, 310-1, 310-N) showing at least scale factors and arithmetic model information and quantized data (330-0, 330-1, 330-N), in order of creating the layers in the set bit groups by analyzing the bit significance by composing the bit groups, from higher significance bits to lower significance bits ; - an inverse quantization portion (410) for restoring decoded scale factors.